New Gemini Models: 3.6 Flash, 3.5 Lite, and Cyber Released

Build production AI agents at scale with Google's newest Gemini models. Experience reduced overall costs, faster performance and enhanced reliability.

Google introduced a new suite of Gemini models (3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber) specifically designed to optimize the efficiency, latency, and reliability required for building AI agents at scale. These models represent a strategic shift toward "agentic workflows," balancing high-quality output with significant reductions in token consumption and operational costs.

Key advancements include:

Gemini 3.6 Flash: A "workhorse" model offering superior coding and multimodal performance with a 17% reduction in output token usage compared to its predecessor.
Gemini 3.5 Flash-Lite: The fastest model in the 3.5 series, capable of 350 output tokens per second, designed for high-throughput tasks like agentic search and document processing.
Gemini 3.5 Flash Cyber: A specialized model integrated into the CodeMender agent infrastructure, focused on identifying and remediating cybersecurity vulnerabilities.
Next-Generation Development: Testing is underway for Gemini 3.5 Pro, and pre-training has commenced for Gemini 4.

Gemini 3.6 Flash: Enhanced Efficiency and Quality

Gemini 3.6 Flash is positioned as the primary model for complex knowledge work and coding. It addresses developer needs for lower latency and improved token efficiency.

Performance and Efficiency Metrics

Token Efficiency: According to the Artificial Analysis Index, 3.6 Flash consumes 17% fewer output tokens than 3.5 Flash. In specific benchmarks like DeepSWE by Datacurve, token reduction reaches up to 65%.
Cost-Effectiveness: The model is priced lower than 3.5 Flash at $1.50 per 1 million input tokens and $7.50 per 1 million output tokens.
Coding and Precision: It delivers higher precision with fewer unwanted edits and execution loops. In the DeepSWE benchmark, it achieved 49% compared to the 37% managed by the 3.5 Flash model.
Computer Use: Performance in OSWorld-Verified tasks improved to 83.0% (from 78.4%). Computer use is now a built-in client-side tool.

Benchmark	Gemini 3.6 Flash	Gemini 3.5 Flash
DeepSWE	49%	37%
MLE Bench (ML Research)	63.9%	49.7%
OSWorld-Verified	83.0%	78.4%
GDPval-AA v2	1421	1349

Gemini 3.5 Flash-Lite: High-Throughput Scaling

Designed for low-latency tasks where high volume is critical, 3.5 Flash-Lite is optimized for agentic search and massive data processing.

Technical Capabilities

Speed: Recognized as the fastest model in the 3.5 series, delivering 350 output tokens per second.
Pricing: Highly competitive at $0.30 per 1 million input tokens and $2.50 per 1 million output tokens.
Thinking Levels: Developers can configure the model to prioritize low-latency execution (minimal/low thinking levels) or engage higher thinking levels for multi-step subagent workloads.
Comparative Advantage: 3.5 Flash-Lite outperforms 3 Flash on several metrics, including SWE-Bench Pro (54.2% vs. 49.6%) and OSWorld-Verified (74.0% vs. 65.1%).

Application Strengths

E-commerce: Extracting and synthesizing product features from massive datasets.
Web Design: Generating unique design concepts instantly when working as a subagent to 3.6 Flash.
Translation and Summarization: Multimodal understanding for scaling tasks such as receipt translation.

Gemini 3.5 Flash Cyber: Security and Vulnerability Remediation

Gemini 3.5 Flash Cyber is a specialized version of the 3.5 Flash model, fine-tuned specifically for cybersecurity applications.

CodeMender Integration

The model operates within CodeMender, an agent infrastructure that utilizes multiple 3.5 Flash Cyber agents working in orchestration to detect, validate, and patch code security issues.

Performance: Achieves competitive "frontier" performance on the CyberGym benchmark.
Operational Intent: Aimed at allowing "frontline defenders" to fix critical vulnerabilities before they are exploited.

Deployment and Safety

Due to the "dual-use nature" of cybersecurity technology, Google has implemented a controlled deployment strategy:

Limited Access: Available exclusively to governments and trusted partners via a pilot program.
Safety Safeguards: The 3.6 Flash model also includes Frontier Safety safeguards against Chemical, Biological, Radiological, and Nuclear (CBRN) misuse and cyber-offense. These upgrades increase resistance to jailbreaks while minimizing refusals for beneficial use-cases.

Future Roadmap and Availability

Model Pipeline

Gemini 3.5 Pro: Currently in testing with partners; broad availability is expected soon.
Gemini 4: Google has initiated its "most ambitious pre-training run yet" for the next generation of models.

Access Channels

The new models are available through several platforms:

For Developers: Gemini API via Google AI Studio and Android Studio; 3.6 Flash is also available in Google Antigravity.
For Enterprises: Gemini Enterprise Agent Platform and the Gemini Enterprise app.
For General Users: The Gemini app, with 3.5 Flash-Lite also rolling out in Google Search.

My SaaS

Acluebox

Build modular and reusable system prompts with my SaaS,

Acluebox

. Also, free prompt template generators there.